In a new study, researchers at North Carolina State University had 28 high school students create their own machine learning artificial intelligence (AI) models to analyze data. The goal of this project was to help students explore the challenges, limits, and possibilities of AI, and to ensure that the workforce of the future is ready to take advantage of AI tools.
The survey was conducted in conjunction with a high school journalism class in the Northeast. Since then, researchers have expanded the program to high school classrooms in multiple states, including North Carolina. Researchers at North Carolina State University are considering partnering with other schools to work together to bring the curriculum into the classroom.
Shiyan Jiang, lead author of the paper and assistant professor of design and technology at North Carolina State University, said: “We want our students to be aware of the possibilities and challenges of AI, and to think about how the next generation of themselves can respond to the evolving role of AI and society. We want to prepare our students for the workforce.”
For this study, researchers developed a computer program called StoryQ that allows students to build their own machine learning models. The researchers then conducted teacher workshops on machine learning curricula and technology, where he held weekly one-and-a-half hour sessions for a month. For teachers who signed up to participate further, the researchers recapitulated the participating teachers’ curriculum and worked out the logistics.
“We created StoryQ technology to allow students in a high school or undergraduate classroom to build what we call a ‘text classification’ model,” said Jiang. “We wanted to lower the barrier so students could really see what was going on with machine learning instead of struggling with coding. We created StoryQ, a tool that helps you understand the nuances of building models.”
Teachers who opted in led a journalism class over a 15-day lesson and used StoryQ to evaluate a series of Yelp reviews about ice cream parlors. Students developed a model that predicted whether a review would be ‘positive’ or ‘negative’ based on language.
“Teachers saw the relevance of the program to journalism,” says Jiang. “This was a very diverse class, with many underrepresented students in the STEM and computing fields. We had a great discussion about.”
Researchers found that students were making hypotheses about certain words in Yelp reviews. We thought that this would allow us to predict whether the reviews would be positive or negative. For example, I expected reviews containing the word “Like” to be positive. The teacher then asked the students to analyze whether the model correctly classified the reviews. For example, students who used the word “like” to predict reviews found that more than half of the reviews containing that word were actually negative. The researchers then said the students tried to improve the model’s accuracy through trial and error.
“Students learned how these models make decisions, the role humans can play in creating these technologies, and the kinds of perspectives that can be captured when creating AI technologies.” Jiang said.
From their discussions, the researchers found that students reacted differently to AI technology. For example, students were deeply concerned about the possibility of using AI to automate the process of selecting students or candidates for opportunities such as scholarships and programs.
For future classes, the researchers created a shorter five-hour program. They started the program at his two high schools in North Carolina, as well as schools in Georgia, Maryland, and Massachusetts. The next phase of research will explore how teachers from different disciplines can work together to launch AI-focused programs and create communities of AI learning.
“We want to expand our implementation in North Carolina,” Zhang said. “If there are schools that are interested, we are always ready to bring this program into their schools. We know teachers are very busy, so we offer shorter professional development courses and even scholarships for teachers.” We provide: join and teach classrooms as needed, or show teachers how to teach the curriculum so they can replicate, adapt, and modify it.We support teachers in any way we can. To do.”
The survey, “Data Modeling Practices and Processes for High School Students: From Modeling Unstructured Data to Evaluating Automated Decision Making,” was published online in the journal March 13. Learning, Media, Technology. Co-authors include Hengtao Tang, Cansu Tatar, Carolyn P. Rosé, and Jie Chao. This work was supported by the National Science Foundation under grant number 1949110.